Meindert’s HPD315 project

History Thirty years ago I bought a pair of HPD 315 speaker and build them into the Devon cabinet, one of the

enclosures in the construction guide of Tannoy. Ten years later I measured the TS-parameters and build,

using the original Small AES-paper, a 100 liter cabinet, a heavy triangular column 100 cm high. But, after a

while I changed to two B&W801 F’s, I sold a few years later because they didn’t sound good at all (too much

distortion in mid and high, and a thick bass). After I build the SEAS Odin MK3 (quite good), it was due to get

my Tannoy’s from the attic, repair the roll surround (see elsewhere) and build a new cabinet using the

modern measurement and simulation techniques the computer offers today.

The Plan My personal design criteria are:

Active filtering: now easily done digitally with for example miniDSP

Small size, since my listening room is small (lowest fundamental 40Hz)

Simple, light but sturdy construction, to make it easily to build and to transport

The design procedure was as follows:

- Measurement of the frequency & phase response of the sound pressure

- Simulation of the speaker

- Designing of the speaker

- Tuning of the active filter

- Measurement of difference between passive and active filter

Measurement of the frequency & phase response of the sound pressure I used the HobboyBox HBX V.6.5 measuring software with a iSEMcon EMM-8 calibrated microphone on my

DELL Studio laptop with good internal soundcard (24 bits, 48000kHz, THD<-70dB). The measurements were

done on a self made DIN 45575 standard baffle (165 x 135, speaker at 65 & 55 cm) on which I placed the

speaker in a 100 liter closed and moderate damped box.

Figure 1 Measuring a Tannoy HPD 315 on a DIN baffle with calibrated microphone using MLS & time windowing

I used the standard MLS-technique of the measurement software with a 1V noise signal and a 5 msec

measurement time window in order to make possible a semi-free field measurement under a low ceiling (2,6

meter) and a cramped space. The microphone was placed on axis at 1 meter measured from the baffle. The

results in a diagram in HBX V.6.5.0:

The woofer shows an interesting response: there is a dip between 1000 and 2000 Hz and has a steep peak at

2200 Hz. The response is only useful till 1000Hz. The waterfall spectrum shows like this:

Note the ringing of the woofer at a frequency around 2kHz. Even more interesting is the typical horn tweeter

response:

Figure 4 Amplitude (red) and phase (blue) of a HPD 315 Horn Tweeter on a DIN45575-baffle at 1 meter & 1V RMS

Figuur 2 Amplitude (red) and fase (blue) of a HPD 315 on a DIN45575-baffle at 1 meter & 1V RMS

Figure 3 A waterfall spectrum of a Tannoy HPD 315 Woofer on a DIN-baffle

With HBX you can export with it measurement points to be used as a input for the simulation on the BOXSIM

simulation program from Visaton. The data is set in steps of 150 Hz is:

Frequency Hz Amplitude Woofer dB

Absolute Phase° woofer

Amplitude Tweeter dB

Absolute Phase°

Tweeter 150 84,0317 89,0833 56,5833 99,5905

300 85,0143 51,7021 69,6169 -46,0829

450 83,7141 44,821 79,8983 -83,0904

600 82,6962 39,1852 85,5849 -85,077

750 82,5677 35,2524 87,7585 -81,2497

900 82,591 35,5834 89,8359 -79,5615

1050 81,3322 36,9232 91,7805 -77,0611

1200 78,6538 29,4548 93,4815 -73,5094

1350 78,3203 12,2942 95,208 -68,4004

1500 81,8756 1,9884 96,7101 -60,5908

1650 84,2227 4,0812 97,2044 -52,2396

1800 85,837 10,9202 97,586 -45,0317

1950 86,1785 16,729 97,4913 -38,8633

2100 88,5742 25,4718 96,9107 -36,2472

2250 89,7168 43,0225 97,0453 -37,0036

2400 88,7109 61,9113 98,317 -35,972

2550 85,7669 72,6124 99,5719 -29,9004

2700 83,9098 73,9087 99,9861 -20,7375

Figuur 5 A waterfall spectrum of a Tannoy HPD 315 Tweeter on a DIN-baffle

Frequency Hz Amplitude Woofer dB

Absolute Phase° woofer

Amplitude Tweeter dB

Absolute Phase°

Tweeter 2850 83,6772 74,3308 99,4905 -12,0265

3000 83,6293 78,2233 98,1774 -7,5593

3150 83,1959 83,8544 97,3375 -8,0581

3300 82,9866 91,5064 97,4184 -9,8615

3450 81,6866 100,1659 98,2589 -8,6886

3600 79,3604 103,5801 98,6546 -3,8065

3750 78,5478 103,5123 98,1696 0,941

3900 78,4081 107,4856 97,5518 2,9738

4050 76,9164 113,3141 97,3372 3,3799

4200 75,1397 116,3832 97,4868 4,1138

4350 73,497 117,4537 97,9334 6,9148

4500 71,7648 117,1228 97,9711 11,5734

4650 69,9131 114,4115 97,6858 16,2764

4800 68,6014 110,2524 97,384 21,0257

4950 67,2475 105,6449 96,671 25,4272

5100 65,8238 98,8228 95,8736 28,5836

5250 65,3872 91,4816 95,1496 31,2656

5400 65,0991 86,1708 94,1273 33,3324

5550 64,7069 81,8588 92,9533 33,6477

5700 64,57 78,8253 91,8748 32,1066

5850 64,1134 77,3318 91,0827 29,792

6000 63,2007 75,7445 90,2017 27,0211

6150 61,4966 70,496 89,1392 22,0637

6300 60,1679 58,7908 88,746 15,2792

6450 61,2898 47,8735 89,0452 9,817

6600 62,2416 42,9421 89,3748 6,8913

6750 63,1211 41,7587 89,5371 5,7363

6900 63,511 43,3633 89,2637 4,7462

7050 63,193 45,4201 88,8589 2,4195

7200 62,2987 45,1834 88,6412 -0,9815

7350 61,5247 41,7033 88,5732 -4,7768

7500 61,4996 37,7074 88,6789 -8,7625

7650 61,422 34,9812 89,245 -11,5652

7800 61,0398 31,5345 89,8831 -11,3363

7950 61,07 26,8961 89,6327 -9,9938

8100 61,7614 23,8659 88,6199 -12,4838

8250 62,4246 24,2376 88,3717 -19,2505

8400 62,1968 24,8381 89,201 -25,9045

8550 62,4358 24,4393 90,8146 -28,2798

Frequency Hz Amplitude Woofer dB

Absolute Phase° woofer

Amplitude Tweeter dB

Absolute Phase°

Tweeter 8700 62,9202 25,725 92,2415 -24,9576

8850 63,4012 29,9761 92,5775 -19,2216

9000 63,1877 36,4102 92,414 -14,4329

9150 61,9637 41,3318 91,9796 -11,3131

9300 60,3804 41,4255 91,6343 -9,2865

9450 59,6126 38,0243 90,9515 -8,8574

9600 59,8219 35,9939 90,8422 -9,3225

9750 59,9772 37,2331 90,8318 -8,1994

9900 60,0041 40,9569 90,2467 -6,7537

10050 59,7117 46,907 89,4423 -6,8963

10200 58,6952 53,3921 88,3485 -9,995

10350 57,4516 59,3571 87,7367 -16,1047

10500 55,9586 65,746 87,5072 -24,4849

10650 53,9236 73,3452 88,9309 -30,9339

10800 50,3682 80,7612 90,6001 -30,5642

10950 42,3233 70,8506 91,4012 -25,3115

11100 35,5503 24,9625 91,2003 -19,3412

11250 44,7593 -6,8303 90,2428 -16,1428

11400 46,7577 -2,1598 89,4341 -16,2344

11550 46,3594 4,0792 88,9312 -17,4402

11700 44,6822 7,5581 88,0909 -20,2849

11850 41,8909 6,4703 87,7554 -25,2415

12000 36,3406 -11,5804 88,0941 -29,4492

12150 35,2147 -47,6217 88,6642 -30,5283

12300 42,3031 -65,4388 88,4854 -29,911

12450 45,562 -52,6158 87,5989 -31,8224

12600 43,5266 -34,0435 86,8543 -38,4268

12750 36,488 -33,0963 87,5667 -45,6444

12900 27,2124 -75,7796 88,6492 -48,7717

13050 35,9228 -113,798 89,1827 -49,1183

13200 36,2942 -121,506 89,106 -50,2542

13350 35,1647 -140,452 88,9721 -54,372

13500 37,4901 -164,525 89,504 -60,7211

13650 42,0076 -178,031 91,2849 -64,9382

13800 45,7887 -178,197 93,4239 -63,1184

13950 47,4365 -172,794 94,9317 -56,2123

14100 47,5004 -170,445 95,8609 -46,2272

14250 47,4554 -174,358 95,5687 -36,1687

14400 48,7308 -180,491 94,8037 -28,8348

Frequency Hz Amplitude Woofer dB

Absolute Phase° woofer

Amplitude Tweeter dB

Absolute Phase°

Tweeter 14550 51,0721 -181,792 93,6115 -25,4296

14700 52,856 -176,412 92,9931 -25,0477

14850 53,1332 -169,545 92,3822 -26,9891

15000 53,1251 -164,175 93,5463 -26,3774

15150 52,4328 -160,746 93,4416 -22,0638

15300 51,4511 -161,182 93,3942 -18,2872

15450 50,7956 -165,469 93,5688 -12,5537

15600 51,486 -168,576 92,5868 -6,8813

15750 51,7379 -167,479 91,7708 -3,6601

15900 51,0207 -165,945 91,3651 0,6074

16050 49,7099 -167,005 89,6558 3,6587

16200 47,4755 -174,687 88,3303 2,5826

16350 46,0291 -191,345 87,8146 1,8458

16500 47,1259 -209,876 86,6559 2,0391

16650 49,5191 -222,521 84,9422 -1,1274

16800 52,5591 -226,231 83,881 -7,9472

16950 54,6811 -221,946 83,5684 -15,3615

17100 55,3162 -215,55 83,9219 -20,7377

17250 55,3876 -210,854 84,5737 -21,812

17400 55,361 -207,48 84,427 -19,6742

17550 55,1549 -204,758 83,4042 -17,9311

17700 54,6821 -203,543 80,6201 -25,5337

17850 55,0625 -200,794 81,6984 -37,2335

18000 54,3642 -195,629 83,594 -39,0459

18150 52,9153 -192,186 83,9358 -35,0158

18300 51,1073 -191,356 83,9898 -29,3323

18450 48,3348 -195,733 82,4636 -25,2588

18600 44,9677 -212,813 80,6198 -28,3458

18750 46,9287 -229,999 80,8457 -33,2493

18900 48,5063 -232,711 82,4939 -27,2686

19050 47,0648 -233,546 79,7701 -18,0867

19200 45,7229 -240,601 75,1173 -26,9577

19350 45,0465 -249,758 77,428 -35,4932

19500 43,7318 -263,022 75,558 -35,7624

19650 45,5384 -274,673 74,2924 -44,4212

19800 48,0363 -271,381 78,4803 -38,4685

19950 47,3961 -256,198 77,0377 -11,4954

Simulation of the speaker In order to use the simulator properly, the simulation with Boxsim had to be calibrated first: the simulation

of the speaker build into the DIN-baffle had to give the same result as the measurement. The calibration

failed because of a few reasons (my informed guess): the actual passive filter I used was not the same as the

simplified filter I build into the simulator, and/or the acoustical centers of the speaker where not right (I did

not measure them). I experimented with different distances between the acoustical centers and I filter I

found on Hilberinks site, but I did not succeed in simulating the actual passive filtered speaker response.

So I took another route: I wanted to know the simulated effect of the enclosure on the response and than

choose the one with the smallest effect. Hence I loaded Boxsim with measuring points of an ideal speaker

woofer and tweeter with the bandwidth from 50 to 12000Hz. I filtered them actively at 1000 Hz and got the

best results with baffle of 80 x 50, the speaker placed 20 from the top and the baffle ends tapered.

Figure 6 Effect of the cabinet dimensions on a ideal loudspeaker response at the optimal box dimensions of about 80 x 50 cm

As you can see, the box dimensions and the active filter result into a lift between 500 and 1000 Hz, a dip

between 1000 and 2000Hz and a ripple beyond 2000Hz. The dip is especially nasty since the speaker itself

has an dip in this region, as you will see below. I will discuss this problem later on when I show the active

filter tuning. In the end I chose a cabinet of 63 x 41 x 32, for various reasons a explain in a minute, that gives

the following simulated response of figure if equipped with ideal speakers:

Figuur 7 Effect of the cabinet dimensions on a ideal loudspeaker response with baffle dimensions of 63 x 41 cm

Shaping the box In choosing the exact dimensions I took in to account the following:

- The tweeter should be at ear height, that is 90 cm, so a small box on a standard or a big stand alone

box should work

- The cut off frequency of the speaker should be 40 Hz or higher, since the fundamental frequency of

my listening room is 40 Hz. Choosing a lower frequency is useless. If I want a speaker with ‘slam’ I

don’t need deeper basses, but a optimal timing between woofer and tweeter, so to produce shock

waves in my room.

The next two design criteria need some extra explanation:

A fast or a deep bass tuning? Normally a Tannoy is build within a bass reflex box. This gives more bass in the frequency domain, but is

costs speed in the time domain, since a bass reflex box is in principle a box with a low frequency gong build

into it. A closed box design of 40 liter reacts fast (with Unified Box Model 408):

A optimal Butterworth bass reflex design of 100 liter reacts slower:

I still chose a bass reflex design of 70 liter with a cut off frequency of 40 Hz. If I wanted it to sound faster I

could simply fill the ports, giving the following approximated 70 liter closed box response:

Non-resonant internal dimensions But what exact internal dimensions should the box have? There should be as little as possible standing waves

inside the box. So I calculated half the length of an standing wave of all the notes of western music. The

internal dimensions should be somewhere between these lengths:

Frequency 247 262 279 294 311 330 349 370 392 415 440 466 494 523 558

Half wave length 0,695 0,656 0,615 0,584 0,552 0,521 0,491 0,464 0,438 0,413 0,390 0,368 0,347 0,328 0,307 Optimal dimension 0,675 0,635 0,599 0,568 0,536 0,506 0,478 0,451 0,425 0,402 0,379 0,358 0,338 0,318 0,300

The internal dimensions chosen are 63,5 x 38 x 33,8 cm. At these dimensions also checked the higher order

resonances of the longer dimension at frequencies other than the western tuning, would not coincide with

the resonances in a shorter dimension. In the end the cabinet panels appeared remarkably silent.

Chipboard or MDF? Normally you would take thick MDF to build a box. This would get very heavy, too heavy. A better choice is a

lighter but stronger cabinet using a internal board reinforcement, making the front baffle acoustically

practically dead. I choose 16 mm chipboard panels for a calculated weight of 18kg (with speaker) and one

internal board reinforce just below the speaker, damping of every bending waves coming from it and .

Chipboard? Yes, chipboard is better. I tested the vibration sound of both a chipboard and a MDF 12 mm 80 x

50 cm panel in an closed box with a small speaker inside(green line). The result, chipboard (red) won:

Tuning the active filter

After assembling the cabinet with the speaker and placing it in a measurement setting, I started tuning the

active filter via the USB link to a miniDSP active filter. Using a interfacing and communication software the

‘TWO WAY CROSSOVER PEQ’ can be tuned as required:

Figuur 8 Sound pressure 50 cm from 12 mm MDF (blue) and chipborard (red) panel

For this tuning some knowledge of filtering techniques is required, since filters interact on mysterious ways if

you haven’t got the knowledge. This time I use HOLMImpuls software to measure the results.

The problem

This response is not easily to be filtered. The woofer rolls of at 1000Hz, the cut off frequency set by Tannoy’s

passive filter, but after 1300 goes up and down again with a peak at 2200Hz, leaving us with a gap to fill with

.. sound from the tweeter. The tweeter has a typical horn response to be made flat. What to do?

The solutions I adopted a principle in filtering: Ockham’s razor, or: less is more. Then, first the simplest part: the tweeter

rolls off at low frequency even without any filtering added. So I added only a real life capacitor of 30 uF

(2x15) only to protect the tweeter and giving it a 1st order high pass filter set at 500 Hz, the resonance filter

of the tweeter (it has resonance frequencies at about 2000, 8000 and 14500Hz ... ).

Second, I gave the woofer a second order low pass Bessel filter at 1000Hz. Bessel is a very moderate filter,

resembling a 2nd order Linkwitz Riley Filter (the LW rolls off faster, too fast here). More filter options are

possible, but I chose Bessel because of its moderate nature in the time domain. In the next step the tweeter

response is flattened out with a step filter: -9 dB flat from 5300Hz downwards (Q=3). The result:

In the high frequencies there are a few resonance peaks to dealt with later, but first the deep dip between

1000 and 2600Hz. There are a lot of options, two simple ones: lifting this regions with the parametric

EQualiser or fiddling with a forgotten parameter: time delay, either of the tweeter, or the woofer. The time

delay needed depends on the physical dimensions of the woofer/tweeter combination and the phase of

both woofer and tweeter driven with a certain filter in the cross over region, so trial and error is the easiest

way to optimize the results. Delaying the woofer (not the tweeter) between 0,2 – 0,3 msec (7 – 10 cm;

depending on the type of woofer filtering used) did the trick (blue line).

Last but not least: supressing the (resonance) peaks and checking the phase (dotted line):

An alternative solution with 3de order Butterworth filtering looks quite similar:

The response bellows up and down between 1 and 5 kHz. I couldn’t hear the difference between Bessel and

Butterworth filtering, yet. A could manage to get it flat with the parameticequaliser, but I choose not to: less

is more.

Is active filtering better? I did a comparison of the original passive filter with a slithly different active filter tuning:

The top red line is the original filter, the top blue line the active filter: the active filtered is less uneven.

Below a indication of the distortion of the speaker: the active filtered speaker is a bit better: approximately -

46 dB (0,5%) THD between 2 and 5kHz, less than 1% above 800Hz. Not bad for a 30 year old speaker. The

results could get better if my measuring conditions would have been better (the distortion measurements

does not use time windowing, so panel vibrations from my room could influence the measurements).

What’s up next? After this project I would like to try and check a few things:

- Simulate the speaker using LSPcad from Ijdata, like all big speaker builders do

- Measure the psycho-acoustic differences between different filter settings

- Test the psycho-acoustic effect of a bigger baffle (50 x 80) (need a bigger house first)

- using a automatic filter optimalisation routine I would like to tune the digital filter so to get a

Linkwitz-Riley response on both the woofer and tweeter measured sound pressure

Meindert Scholma, Amsterdam, 2011